A cloud platform computing power resource allocation system and method based on balanced load

Abstract

This invention discloses a cloud platform computing power resource allocation system and method based on load balancing, relating to the field of computing power resource allocation technology. The application includes: constructing a computing power supply and demand matching platform, connecting with idle computing power resource providers, and receiving computing power requests from computing power resource lessors; based on the computing power requests from the lessors, determining the cluster time, and allocating idle computing power resources to the lessors in the cluster; the lessors setting up key communication, and after key verification, the system accessing the cloud platform of the allocated lessor, achieving intelligent decision-making and optimizing the scheduling of allocated idle computing power resources in the cloud platform. This invention can maximize overall resource utilization, reduce operating costs, minimize task execution latency, and improve user experience. Simultaneously, dynamic allocation based on computing power sharing and cluster processing according to different time nodes improves the overall utilization efficiency of computing power resources.

Description

Technical Field

[0001] This invention relates to the field of computing power resource allocation technology, specifically a cloud platform computing power resource allocation system and method based on load balancing. Background Technology

[0002] Computing power, the ability of information technology equipment to process data and perform calculations, is a new type of productive force following heat and electricity, and has become a key element driving technological innovation and industrial transformation. However, in the current allocation of computing power, at the product level, the domestic computing power industry chain is fragmented, and an industrial ecosystem has not yet been formed; at the spatial level, the eastern region is economically developed and has a concentration of technology companies, resulting in strong demand for computing power but insufficient supply, while the central and western regions suffer from a surplus of computing power resources, indicating an unreasonable allocation of regional resources.

[0003] Therefore, developing demand-side computing infrastructure based on the core concept of computing power sharing is key to solving the current problem of inefficient utilization of computing resources. However, regarding the concept of computing power sharing, how to achieve sharing, how to effectively integrate dispersed and underutilized computing resources, and how to achieve more flexible supply and demand matching through fine-grained task scheduling are all unresolved issues. Furthermore, after computing power is allocated, how each allocated cloud platform can achieve internal supply and demand adjustment, how to solve the resource utilization inefficiency caused by static allocation, and how to achieve fine-grained, dynamic load balancing to improve the efficiency and reliability of cloud platforms are also currently unresolved problems. Summary of the Invention

[0004] The purpose of this invention is to provide a cloud platform computing resource allocation system and method based on load balancing to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a cloud platform computing resource allocation method based on load balancing, the method comprising:

[0006] Build a computing power supply and demand matching platform to connect with idle computing power resources, acquire idle computing power resources, open the supply and demand matching portal, and receive computing power requests from computing power resource lessors;

[0007] Based on the computing power requests of the computing power resource lessees, the cluster time is determined, and a computing power sharing instruction is issued based on the cluster time to allocate idle computing power resources to the computing power resource lessees in the cluster.

[0008] The computing resource lessee sets up key communication. After key verification, the system connects to the cloud platform of the allocated computing resource lessee to realize intelligent decision-making and optimize the scheduling of idle computing resources allocated in the cloud platform.

[0009] According to the above technical solution, the construction of a computing power supply and demand matching platform to connect with parties with idle computing power resources and obtain idle computing power resources includes:

[0010] The idle computing resources are registered and accessed by the parties, and a proxy program is provided to the computing power supply and demand matching platform. The computing power supply and demand matching platform obtains idle computing resources based on the proxy program.

[0011] The computing power supply and demand matching platform sets up resource monitoring and reporting ports to monitor the server status information data of idle computing power resources in real time. The server status information data includes: utilization rate, temperature, power consumption and network status.

[0012] The computing power supply and demand matching platform is also equipped with resource isolation and security sandboxes. It uses containerization technology to isolate each task of the computing power resource lessee and has the ability to restrict the network access and file system read and write permissions of the task.

[0013] According to the above technical solution, the opening of the supply and demand matching portal to receive computing power requests from computing power resource lessors includes:

[0014] Verify the validity of the identity information data of the computing power resource lessee based on the gateway or authentication service;

[0015] After the identity information data is verified, it is confirmed whether the computing power request of the computing power resource lessee matches the cost paid.

[0016] When the computing power request of the computing power resource lessee matches its payment cost, the verification is successful, and the computing power request is forwarded to the backend service and converted into a scheduling node under the unified format of the platform; each scheduling node contains the request access time, computing power request amount, estimated computing power usage time, and computing power usage address.

[0017] The scheduling nodes are merged into the sorting queue of the computing power supply and demand matching platform.

[0018] According to the above technical solution, determining the cluster time includes:

[0019] First restriction order;

[0020] Identify the idle computing resources data of each idle computing resource provider and form a set. ,in, The idle computing resources data represent n parties with idle computing resources respectively. The set is traversed and the idle computing resources data of any two or more parties with idle computing resources are combined to form computing power combined data. The computing power combined data is added to the set. Only one of the same computing power combined data is kept to form the first data set.

[0021] Based on the sorting queue of the computing power supply and demand adaptation platform, the computing power request amount in the scheduling node is continuously added sequentially. If the computing power request amount in the scheduling node that is continuously added at a certain moment is equal to any data in the first data set, the addition is stopped and that moment is determined as the cluster moment.

[0022] According to the above technical solution, determining the cluster time also includes:

[0023] Second restriction order;

[0024] Set a minimum resource waste threshold, and obtain data from historical cluster data where resource waste is lower than the minimum resource waste threshold. The resource waste refers to the product of the difference between all idle computing resources of the allocated computing resource idle party and the computing resource actually used by the computing resource lessee and the usage time.

[0025] In data where resource waste is below the minimum resource waste threshold, the number of computing resource lessees and their maximum waiting time are extracted from the cluster data. Based on resource waste, the number of computing resource lessees, and their maximum waiting time, a three-dimensional coordinate scatter plot is formed, and a functional relationship is constructed:

[0026] The function type is determined based on the distribution of three-dimensional coordinate scattered points. The function type includes linear model and polynomial model. If the three-dimensional coordinate scattered points are almost linearly distributed, the linear model is used; otherwise, the polynomial model is used.

[0027] We define a function F as the prediction function, and input the 3D coordinate scatter points into the prediction function to form the difference between the true function and the predicted function. Then in When the value is minimized, the predicted function F is closest to the true function relationship, where m refers to the number of scatter points in the three-dimensional coordinate system;

[0028] The least squares method is used based on the prediction function F, in When the value is minimized, the parameters are solved.

[0029] The output function relationship based on solving the parameters is used as the true function relationship;

[0030] The system sets a threshold and sorts the data in the computing power supply and demand matching platform based on the scheduling nodes. The data in the queue is then input into the real function relationship. If the output value is less than the system-set threshold, the moment is determined to be a cluster moment. The resource waste in the data in the queue is the minimum difference between the computing power request amount in the scheduling nodes that is continuously added at a certain moment and any data in the first data set.

[0031] According to the above technical solution, when the system receives the feedback of the first restriction instruction or the second restriction instruction, it issues a computing power sharing instruction. At this time, the computing power resource lessees in the queue are formed into a cluster for processing, and the queue is reopened.

[0032] According to the above technical solution, the computing resource lessee sets up key communication. After key verification, the system connects to the cloud platform of the allocated computing resource lessee to achieve intelligent decision-making and optimize the scheduling of idle computing resources allocated in the cloud platform, including:

[0033] Deploy lightweight monitoring agents on each physical or virtual computing node of the computing resource lessee to collect real-time performance metrics of the nodes. All monitoring agents will report the collected data to a central state aggregator in real time. The aggregator will clean, summarize and store the data to form a real-time global resource view of a single computing resource lessee.

[0034] The system receives new tasks submitted by computing resource lessees and parses the attached resource requirement description file to obtain the task's explicit requirements for computing, memory, and storage resources.

[0035] Intelligent decision-making based on load balancing and resource compatibility includes: prioritizing the node with the lowest current load rate, and calculating the weighted load values ​​of CPU, memory, and I / O; prioritizing the node that can meet the requirements and has the least resource fragmentation; and prioritizing the scheduling of frequently communicating tasks to the same rack, the same physical machine, or the area with the lowest network latency to reduce network transmission overhead.

[0036] A cloud platform computing resource allocation system based on load balancing, the system comprising:

[0037] The computing power supply and demand matching module is used to build a computing power supply and demand matching platform, connect with idle computing power resources, obtain idle computing power resources, open the supply and demand matching portal, and receive computing power requests from computing power resource lessors.

[0038] The computing power sharing cluster module determines the cluster time based on the computing power request of the computing power resource lessee, issues computing power sharing instructions based on the cluster time, and allocates idle computing power resources to the computing power resource lessee in the cluster;

[0039] The computing resource allocation module is used by the computing resource lessee to set up key communication. After the key is verified, the system connects to the cloud platform of the allocated computing resource lessee to realize intelligent decision-making and optimize the scheduling of the allocated idle computing resources in the cloud platform.

[0040] According to the above technical solution, the computing power supply and demand adaptation module further includes:

[0041] The idle computing resources are registered and accessed by the parties, and a proxy program is provided to the computing power supply and demand matching platform. The computing power supply and demand matching platform obtains idle computing resources based on the proxy program.

[0042] The computing power supply and demand matching platform sets up resource monitoring and reporting ports to monitor the server status information data of idle computing power resources in real time. The server status information data includes: utilization rate, temperature, power consumption and network status.

[0043] The computing power supply and demand matching platform is also equipped with resource isolation and security sandboxes. It uses containerization technology to isolate each task of the computing power resource lessee and has the ability to restrict the network access and file system read and write permissions of the task.

[0044] According to the above technical solution, the computing power supply and demand adaptation module further includes:

[0045] Verify the validity of the identity information data of the computing power resource lessee based on the gateway or authentication service;

[0046] After the identity information data is verified, it is confirmed whether the computing power request of the computing power resource lessee matches the cost paid.

[0047] When the computing power request of the computing power resource lessee matches its payment cost, the verification is successful, and the computing power request is forwarded to the backend service and converted into a scheduling node under the unified format of the platform; each scheduling node contains the request access time, computing power request amount, estimated computing power usage time, and computing power usage address.

[0048] The scheduling nodes are merged into the sorting queue of the computing power supply and demand matching platform.

[0049] Compared with existing technologies, the beneficial effects of this invention are as follows: By optimizing and learning from historical task and resource allocation records, this invention maximizes overall resource utilization and reduces operating costs; minimizes task execution latency, improves user experience and processing efficiency for high-concurrency businesses; multi-dimensional and fine-grained real-time scheduling decisions greatly optimize resource packaging strategies, reduce resource fragmentation, and qualitatively improve overall resource utilization, directly reducing the TCO of the data center; and based on real-time load balancing strategies and performance affinity scheduling, it ensures that tasks are always allocated to the most suitable nodes for rapid execution, significantly shortening task queuing time and execution time, especially in high-concurrency scenarios, resulting in a greatly improved user experience.

[0050] At the application level, flexible scheduling strategies can integrate different types of computing resources onto the computing power sharing platform to alleviate the pressure of high-end computing power shortages. Furthermore, based on computing power sharing, resources are dynamically allocated according to users' actual needs, and cluster processing is performed at different time points to meet the requirements of various application scenarios and improve the overall utilization efficiency of computing resources. Attached Figure Description

[0051] Figure 1 This is a schematic diagram illustrating the steps of a cloud platform computing resource allocation method based on load balancing according to the present invention. Detailed Implementation

[0052] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Example: Figure 1 As shown, this invention provides a cloud platform computing resource allocation method based on load balancing, the method comprising:

[0054] Build a computing power supply and demand matching platform to connect with idle computing power resources, acquire idle computing power resources, open the supply and demand matching portal, and receive computing power requests from computing power resource lessors;

[0055] The construction of the computing power supply and demand matching platform, connecting with parties with idle computing power resources, and acquiring idle computing power resources includes:

[0056] The idle computing resources are registered and accessed by the parties, and a proxy program is provided to the computing power supply and demand matching platform. The computing power supply and demand matching platform obtains idle computing resources based on the proxy program.

[0057] The computing power supply and demand matching platform sets up resource monitoring and reporting ports to monitor the server status information data of idle computing power resources in real time. The server status information data includes: utilization rate, temperature, power consumption and network status.

[0058] The computing power supply and demand matching platform is also equipped with resource isolation and security sandboxes. It uses containerization technology to isolate each task of the computing power resource lessee and has the ability to restrict the network access and file system read and write permissions of the task.

[0059] The open supply and demand matching portal, which receives computing power requests from computing power resource lessors, includes:

[0060] Verify the validity of the identity information data of the computing power resource lessee based on the gateway or authentication service;

[0061] After the identity information data is verified, it is confirmed whether the computing power request of the computing power resource lessee matches the cost paid.

[0062] When the computing power request of the computing power resource lessee matches its payment cost, the verification is successful, and the computing power request is forwarded to the backend service and converted into a scheduling node under the unified format of the platform; each scheduling node contains the request access time, computing power request amount, estimated computing power usage time, and computing power usage address.

[0063] The scheduling nodes are merged into the sorting queue of the computing power supply and demand matching platform.

[0064] Based on the computing power requests of the computing power resource lessees, the cluster time is determined, and a computing power sharing instruction is issued based on the cluster time to allocate idle computing power resources to the computing power resource lessees in the cluster.

[0065] The determination of the cluster time includes:

[0066] First restriction order;

[0067] Identify the idle computing resources data of each idle computing resource provider and form a set. ,in, The idle computing resources data represent n parties with idle computing resources respectively. The set is traversed and the idle computing resources data of any two or more parties with idle computing resources are combined to form computing power combined data. The computing power combined data is added to the set. Only one of the same computing power combined data is kept to form the first data set.

[0068] Based on the sorting queue of the computing power supply and demand adaptation platform, the computing power request amount in the scheduling node is continuously added sequentially. If the computing power request amount in the scheduling node that is continuously added at a certain moment is equal to any data in the first data set, the addition is stopped and that moment is determined as the cluster moment.

[0069] The determination of cluster timing also includes:

[0070] Second restriction order;

[0071] Set a minimum resource waste threshold, and obtain data from historical cluster data where resource waste is lower than the minimum resource waste threshold. The resource waste refers to the product of the difference between all idle computing resources of the allocated computing resource idle party and the computing resource actually used by the computing resource lessee and the usage time.

[0072] In data where resource waste is below the minimum resource waste threshold, the number of computing resource lessees and their maximum waiting time are extracted from the cluster data. Based on resource waste, the number of computing resource lessees, and their maximum waiting time, a three-dimensional coordinate scatter plot is formed, and a functional relationship is constructed:

[0073] The function type is determined based on the distribution of three-dimensional coordinate scattered points. The function type includes linear model and polynomial model. If the three-dimensional coordinate scattered points are almost linearly distributed, the linear model is used; otherwise, the polynomial model is used.

[0074] We define a function F as the prediction function, and input the 3D coordinate scatter points into the prediction function to form the difference between the true function and the predicted function. Then in When the value is minimized, the predicted function F is closest to the true function relationship, where m refers to the number of scatter points in the three-dimensional coordinate system;

[0075] The least squares method is used based on the prediction function F, in When the value is minimized, the parameters are solved.

[0076] The output function relationship based on solving the parameters is used as the true function relationship;

[0077] Taking a linear model as an example, then ,in, Refers to the true functional relationship. Input items Refers to parameter items;

[0078] but When the function is at its minimum, To find the minimum value, we take the partial derivative with respect to the parameter terms and set the equation to zero to solve the problem.

[0079] The system sets a threshold and sorts the data in the computing power supply and demand matching platform based on the scheduling nodes. The data in the queue is then input into the real function relationship. If the output value is less than the system-set threshold, the moment is determined to be a cluster moment. The resource waste in the data in the queue is the minimum difference between the computing power request amount in the scheduling nodes that is continuously added at a certain moment and any data in the first data set.

[0080] When the system receives the first or second restriction instruction, it issues a computing power sharing instruction. At this time, the computing power resource lessees in the queue are formed into a cluster for processing, and the queue is reopened.

[0081] The computing resource lessee sets up key communication. After key verification, the system connects to the cloud platform of the allocated computing resource lessee to realize intelligent decision-making and optimize the scheduling of idle computing resources allocated in the cloud platform.

[0082] The computing resource lessee sets up key communication. After key verification, the system connects to the cloud platform of the allocated computing resource lessee to achieve intelligent decision-making and optimize the scheduling of idle computing resources allocated in the cloud platform, including:

[0083] Deploy lightweight monitoring agents on each physical or virtual computing node of the computing resource lessee to collect real-time performance metrics of the nodes. All monitoring agents will report the collected data to a central state aggregator in real time. The aggregator will clean, summarize and store the data to form a real-time global resource view of a single computing resource lessee.

[0084] The system receives new tasks submitted by computing resource lessees and parses the attached resource requirement description file to obtain the task's explicit requirements for computing, memory, and storage resources.

[0085] Intelligent decision-making based on load balancing and resource compatibility includes: prioritizing the node with the lowest current load rate, and calculating the weighted load values ​​of CPU, memory, and I / O; prioritizing the node that can meet the requirements and has the least resource fragmentation; and prioritizing the scheduling of frequently communicating tasks to the same rack, the same physical machine, or the area with the lowest network latency to reduce network transmission overhead.

[0086] This embodiment also includes a cloud platform computing resource allocation system based on load balancing, the system comprising:

[0087] The computing power supply and demand matching module is used to build a computing power supply and demand matching platform, connect with idle computing power resources, obtain idle computing power resources, open the supply and demand matching portal, and receive computing power requests from computing power resource lessors.

[0088] The computing power sharing cluster module determines the cluster time based on the computing power request of the computing power resource lessee, issues computing power sharing instructions based on the cluster time, and allocates idle computing power resources to the computing power resource lessee in the cluster;

[0089] The computing resource allocation module is used by the computing resource lessee to set up key communication. After the key is verified, the system connects to the cloud platform of the allocated computing resource lessee to realize intelligent decision-making and optimize the scheduling of the allocated idle computing resources in the cloud platform.

[0090] The computing power supply and demand adaptation module also includes:

[0091] The idle computing resources are registered and accessed by the parties, and a proxy program is provided to the computing power supply and demand matching platform. The computing power supply and demand matching platform obtains idle computing resources based on the proxy program.

[0092] The computing power supply and demand matching platform sets up resource monitoring and reporting ports to monitor the server status information data of idle computing power resources in real time. The server status information data includes: utilization rate, temperature, power consumption and network status.

[0093] The computing power supply and demand matching platform is also equipped with resource isolation and security sandboxes. It uses containerization technology to isolate each task of the computing power resource lessee and has the ability to restrict the network access and file system read and write permissions of the task.

[0094] The computing power supply and demand adaptation module also includes:

[0095] Verify the validity of the identity information data of the computing power resource lessee based on the gateway or authentication service;

[0096] After the identity information data is verified, it is confirmed whether the computing power request of the computing power resource lessee matches the cost paid.

[0097] When the computing power request of the computing power resource lessee matches its payment cost, the verification is successful, and the computing power request is forwarded to the backend service and converted into a scheduling node under the unified format of the platform; each scheduling node contains the request access time, computing power request amount, estimated computing power usage time, and computing power usage address.

[0098] The scheduling nodes are merged into the sorting queue of the computing power supply and demand matching platform.

[0099] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for allocating computing power resources of a cloud platform based on balanced load, characterized in that: The method includes: Build a computing power supply and demand matching platform to connect with idle computing power resources, acquire idle computing power resources, open the supply and demand matching portal, and receive computing power requests from computing power resource lessors; Based on the computing power requests of the computing power resource lessees, the cluster time is determined, and a computing power sharing instruction is issued based on the cluster time to allocate idle computing power resources to the computing power resource lessees in the cluster. The computing resource lessee sets up key communication. After key verification, the system connects to the cloud platform of the allocated computing resource lessee to realize intelligent decision-making and optimize the scheduling of the allocated idle computing resources in the cloud platform. The determination of the cluster time includes: First restriction order; determining idle computing power resource data of each computing power resource idle party, forming a set wherein, respectively represent the idle computing power resource data of n computing power resource idle parties, traversing the set to combine the idle computing power resource data of any two or more computing power resource idle parties, forming computing power combination data, supplementing the computing power combination data into the set, retaining only one of the same computing power combination data, forming a first data set; Based on the sorting queue of the computing power supply and demand adaptation platform, the computing power request amount in the scheduling node is continuously added sequentially. If the computing power request amount in the scheduling node that is continuously added at a certain moment is equal to any data in the first data set, the addition is stopped and that moment is determined as the cluster moment.

2. The cloud platform computing resource allocation method based on load balancing according to claim 1, characterized in that: The construction of the computing power supply and demand matching platform, connecting with parties with idle computing power resources, and acquiring idle computing power resources includes: The idle computing resources are registered and accessed by the parties, and a proxy program is provided to the computing power supply and demand matching platform. The computing power supply and demand matching platform obtains idle computing resources based on the proxy program. The computing power supply and demand matching platform sets up resource monitoring and reporting ports to monitor the server status information data of idle computing power resources in real time. The server status information data includes: utilization rate, temperature, power consumption and network status. The computing power supply and demand matching platform is also equipped with resource isolation and security sandboxes. It uses containerization technology to isolate each task of the computing power resource lessee and has the ability to restrict the network access and file system read and write permissions of the task.

3. The cloud platform computing resource allocation method based on load balancing according to claim 1, characterized in that: The open supply and demand matching portal, which receives computing power requests from computing power resource lessors, includes: Verify the validity of the identity information data of the computing power resource lessee based on the gateway or authentication service; After the identity information data is verified, it is confirmed whether the computing power request of the computing power resource lessee matches the cost paid. When the computing power request of the computing power resource lessee matches its payment cost, the verification is successful, and the computing power request is forwarded to the backend service and converted into a scheduling node under the unified format of the platform; each scheduling node contains the request access time, computing power request amount, estimated computing power usage time, and computing power usage address. The scheduling nodes are merged into the sorting queue of the computing power supply and demand matching platform.

4. The cloud platform computing resource allocation method based on load balancing according to claim 3, characterized in that: The determination of cluster timing also includes: Second restriction order; Set a minimum resource waste threshold, and obtain data from historical cluster data where resource waste is lower than the minimum resource waste threshold. The resource waste refers to the product of the difference between all idle computing resources of the allocated computing resource idle party and the computing resource actually used by the computing resource lessee and the usage time. In data where resource waste is below the minimum resource waste threshold, the number of computing resource lessees and their maximum waiting time are extracted from the cluster data. Based on resource waste, the number of computing resource lessees, and their maximum waiting time, a three-dimensional coordinate scatter plot is formed, and a functional relationship is constructed: The function type is determined based on the distribution of three-dimensional coordinate scattered points. The function type includes linear model and polynomial model. If the three-dimensional coordinate scattered points are almost linearly distributed, the linear model is used; otherwise, the polynomial model is used. We define a function F as the prediction function, and substitute the 3D coordinate scatter points into the prediction function to form the difference between the true function and the predicted function. Then in When the value is minimized, the predicted function F is closest to the true function relationship, where m refers to the number of scatter points in the three-dimensional coordinate system; The least squares method is used based on the prediction function F, in When the value is minimized, the parameters are solved. The output function relationship based on solving the parameters is used as the true function relationship; The system sets a threshold and sorts the data in the computing power supply and demand matching platform based on the scheduling nodes. The data in the queue is then input into the real function relationship. If the output value is less than the system-set threshold, the moment is determined to be a cluster moment. The resource waste in the data in the queue is the minimum difference between the computing power request amount in the scheduling nodes that is continuously added at a certain moment and any data in the first data set.

5. A cloud platform computing resource allocation method based on load balancing according to claim 4, characterized in that: When the system receives the first or second restriction instruction, it issues a computing power sharing instruction. At this time, the computing power resource lessees in the queue are formed into a cluster for processing, and the queue is reopened.

6. The cloud platform computing resource allocation method based on load balancing according to claim 1, characterized in that: The computing resource lessee sets up key communication. After key verification, the system connects to the cloud platform of the allocated computing resource lessee to achieve intelligent decision-making and optimize the scheduling of idle computing resources allocated in the cloud platform, including: Deploy lightweight monitoring agents on each physical or virtual computing node of the computing resource lessee to collect real-time performance metrics of the nodes. All monitoring agents report the collected data to a central state aggregator in real time. The aggregator cleans, summarizes and stores the data to form a real-time global resource view of a single computing resource lessee. The system receives new tasks submitted by computing resource lessees and parses the attached resource requirement description file to obtain the task's explicit requirements for computing, memory, and storage resources. Intelligent decision-making based on load balancing and resource compatibility includes: prioritizing the node with the lowest current load rate, and calculating the weighted load values ​​of CPU, memory, and I / O; prioritizing the node that can meet the requirements and has the least resource fragmentation; and prioritizing the scheduling of frequently communicating tasks to the same rack, the same physical machine, or the area with the lowest network latency to reduce network transmission overhead.

7. A cloud platform computing resource allocation system based on load balancing, used to implement the cloud platform computing resource allocation method based on load balancing as described in claim 1, characterized in that: The system includes: The computing power supply and demand matching module is used to build a computing power supply and demand matching platform, connect with idle computing power resources, obtain idle computing power resources, open the supply and demand matching portal, and receive computing power requests from computing power resource lessors. The computing power sharing cluster module determines the cluster time based on the computing power request of the computing power resource lessee, issues computing power sharing instructions based on the cluster time, and allocates idle computing power resources to the computing power resource lessee in the cluster; The computing resource allocation module is used by the computing resource lessee to set up key communication. After the key is verified, the system connects to the cloud platform of the allocated computing resource lessee to realize intelligent decision-making and optimize the scheduling of the allocated idle computing resources in the cloud platform.

8. A cloud platform computing resource allocation system based on load balancing according to claim 7, characterized in that: The computing power supply and demand adaptation module also includes: The idle computing resources are registered and accessed by the parties, and a proxy program is provided to the computing power supply and demand matching platform. The computing power supply and demand matching platform obtains idle computing resources based on the proxy program. The computing power supply and demand matching platform sets up resource monitoring and reporting ports to monitor the server status information data of idle computing power resources in real time. The server status information data includes: utilization rate, temperature, power consumption and network status. The computing power supply and demand matching platform is also equipped with resource isolation and security sandboxes. It uses containerization technology to isolate each task of the computing power resource lessee and has the ability to restrict the network access and file system read and write permissions of the task.

9. A cloud platform computing resource allocation system based on load balancing according to claim 8, characterized in that: The computing power supply and demand adaptation module also includes: Verify the validity of the identity information data of the computing power resource lessee based on the gateway or authentication service; After the identity information data is verified, it is confirmed whether the computing power request of the computing power resource lessee matches the cost paid. When the computing power request of the computing power resource lessee matches its payment cost, the verification is successful, and the computing power request is forwarded to the backend service and converted into a scheduling node under the unified format of the platform; each scheduling node contains the request access time, computing power request amount, estimated computing power usage time, and computing power usage address. The scheduling nodes are merged into the sorting queue of the computing power supply and demand matching platform.

Images